Functional MRI Method Development

National Institute Of Mental Health

Investigators

Linked publications, trials & patents

Abstract

The research projects listed are approximately grouped by the themes mentioned in the Goals and Objectives section. They are each led by either a staff scientist or a post doc in SFIM, and have contributing SFIM members listed. Across-echo Functional Connectivity as a quality metric for multi-echo fMRI (Javier Gonzalez-Castillo, Daniel A Handwerker) With BOLD fMRI, several quality assurance (QC) metrics exist (e.g., temporal signal-to-noise ratio (TSNR), test-retest reliability), yet their interpretation and usefulness remain challenging. For example, higher TSNR suggests better data, yet removal of neuronally-driven BOLD signal would also increase it. Also, perfect reliability across sessions can obscure relevant changes in functional connectivity (FC) over time. To overcome this, we propose a new QC metric for multi-echo (ME) fMRI that quantifies the likelihood of data being dominated by BOLD effects. This metric builds on the observation that functional connectivity, when measured using covariance, has distinct behaviors depending on whether data is dominated by non-BOLD (i.e., noise) or BOLD (which includes fluctuations of neural origin) effects. Using a large multi-echo dataset (436 scans), we show our proposed metric can help evaluate the effectiveness of global signal regression at eliminating purely non-BOLD effects, and negative interactions between NORDIC denoising and tedana denoising. We are also working on defining how to make use of this metric to detect individual connections (i.e., edges) whose strength might be inflated due to non-BOLD effects. Realtime fMRI monitoring of Spontaneous Brain Activity (Javier Gonzalez-Castillo, Marly Rubin, Daniel A Handwerker) The biological significance of spontaneous neural activity remains not fully understood. One hypothesis suggests that an aspect of this activity constitutes internal conscious thought. However, using traditional paradigms, it is difficult to study the link between spontaneous brain activity and self-driven covert cognition. Real-time functional MRI (rtfMRI) offers a means to explore this question by allowing researchers to monitor spontaneous activation of brain configurations of interest as they unfold and use these configurations to trigger introspective surveys to interrogate the contents of ongoing cognition. To accomplish this goal, we are developing a new open-source Python package (rt-cog) that integrates with AFNI to accomplish three tasks: 1) online pre-processing of incoming data, 2) matching to a particular brain configuration of interest, 3) administration of in-scanner introspective surveys when a match occurs. All these operations are highly configurable. Importantly, rt-cog leverages multiprocessing to achieve processing speeds of ~0.15 seconds per fMRI volume on an M4 MacBook Pro. We are currently in the process of final testing. Once testing is completed, we will use this infrastructure to study the conscious correlates of spontaneous activation of different brain systems such as the visual network, the motor network and the language network, among others. Characterization of functional network dynamics using Edge Analysis (Josh Faskowitz, Javier Gonzalez-Castillo) As the human brain rests, neurovascular signals from distinct parts of the brain will fall in and out of synchrony. Quantifying the spatial and temporal patterns of these functional phenomena has the potential to inform clinical fMRI applications and to further our understanding of how brain function emerges across time. Recently, we formulated a mathematical approach to track time-varying synchrony, which we termed edge time series, or ETS. Using ETS we can pinpoint moments during a fMRI scan when synchrony is particularly elevated between many parts of the brain, and extract features such as the count and duration of these moments. Mapping these features reveals both overlap with our current understanding of the brain’s functional connectivity but also provides new information about the timescales of different areas’ synchronization. We identified areas, such as the primary visual areas, that exhibit long periods of increased synchrony. Through our work, we underscore the idea that the brain’s synchrony manifests as punctate events, lasting less than five seconds, spread throughout the time course of a scan. Our work was presented at the 2025 Organization for Human Brain Mapping conference and has been submitted as a manuscript to an academic journal. Systemic Vascular Influences on BOLD Dynamics (Burak Akin, Dan Handwerker) This project centers on cerebral and peripheral hemodynamics which interact through shared systemic factors such as heart rate, vascular tone, and blood pressure — all key elements that modulate cerebral blood flow and influence the BOLD (Blood Oxygenation Level Dependent) signal measured in most neuroimaging studies. We utilized an MRI-compatible blood pressure monitor to capture the blood pressure dynamics over time during simultaneous collection of fMRI time series. Blood pressure fluctuates up to 10% even at rest. Systemic blood pressure exerts a delayed but organized influence on cerebral venous structures, reflecting a passive vascular response shaped by intracranial compliance. The paper is targeted for submission this summer, highlighting a potentially distinct periodic pattern in the brain alongside known rhythms such as those captured by peripheral blood oxygenation and volume measures. Multi-echo fMRI methods (Daniel Handwerker, Javier Gonzalez-Castillo, Joshua Dean) We have continued work improving multi-echo fMRI processing methods. This work is paired with developing an open-source software program called tedana so that the work we do is widely accessible to researchers. The impact of this work is increasing with more researchers using multi-echo methods that were developed and improved within SFIM. During the past year, this work has included: (1) Work developing better ways to include information about heart pulsation and respiration within multi-echo fMRI methods, (2) Ways to integrate a novel acquisition method with more echoes called EPTI (in collaboration with Tom Liu at UCSD), (3) Ways to get more stable results (in collaboration with researchers at The Florey), (4) Ways to use multi-echo information to better evaluate if fMRI data processing successfully improves data quality. Functional CSF volume imaging (Laurentius Huber and Stephanie Swegle) Ms. Swegle is specifically working on imaging methods of CSF volume. She has developed a fast multi-echo, multi-inversion 3D-EPI protocol to capture functional changes in CSF, independently from changes in blood volume and BOLD. She found indications that CSF volume changes as a function of brain activation and during drowsiness. She presented this method in the form of a conference talk to other experts in the field. This method might become a useful tool for investigating the mechanism of the glymphatic system in health and disease. Voluntary Saccades Show Robust Layer Activation in Frontal Eye Field and Midbrain Oculomotor Networks (Burak Akin, Sharif Kronemer, Laurentius Huber, Daniel Handwerker) Voluntary saccades require coordinated activity across cortical and subcortical regions involved in motor planning, attention, and executive control. In this study, we employed ultra-high field 7T fMRI with parallel transmission (pTx) and high-acceleration GRAPPA (R=16), covering the whole brain to investigate layer-specific activation patterns during auditory-cued saccades, focusing on fronto-striatal and midbrain structures, including the frontal eye field (FEF), supplementary eye field (SEF), caudate nucleus, and dorsolateral prefrontal cortex (DLPFC). The DLPFC has been implicated in the preparatory control of volitional eye movements, while FEF/SEF and caudate are central to saccade execution. Our preliminary results, based on temporal dynamics and laminar profiles, support a bidirectional model of saccade generation: a top-down pathway originating in the prefrontal cortex projecting to the caudate, and a bottom-up loop from caudate to FEF/SEF, consistent with cortico-striatal-thalamo-cortical circuitry underlying volitional oculomotor control. Processing Methods to Capture Layer Dynamics and Patterns (Burak Akin) While layer fMRI has been shown to have outstanding spatial precision, it currently lacks connectivity tools that can capture local correlation dynamics to highlight layer specific activity. To address this, we developed two methods: patch analysis and layer-specific regional homogeneity, which reveal localized activity patterns—including dynamic properties, directional tendencies (feedforward vs. feedback), and fine spatial specificity—in a purely data-driven manner. These tools, validated on both public datasets and NIH-acquired data, will be integrated with formal mathematical noise model to advance toward a publishable framework. Hippocampus involvement in working memory (Catherine Walsh) Maintenance of information in working memory is fundamental for higher-order cognition; recent theoretical frameworks have suggested that information can be maintained in the face of interruptions through processes mediated by medial temporal lobe structures such as the hippocampus. In this project, we are seeking to empirically test the role of the hippocampus in the maintenance of information over a short delay period. To that end, we have adapted a dual-retrocue working memory task for use in 7T fMRI scanning. Data collection is in progress, with 5 pilot MRI sessions completed using this task. Preliminary univariate analyses suggest that the hippocampus may be involved in refreshing information as it is maintained and retrieving previously task-irrelevant information from a more long-term memory-like state as it becomes task-relevant. We are in the process of developing analyses to further characterize the role of the hippocampus in working memory processes, including identifying category-specific information at each timepoint across trials using machine-learning algorithms, indexing hippocampal-cortical functional connectivity and relating activation and decoding performance in the hippocampus to a post-scan memory test. These results have been submitted for a poster presentation at the 2025 Society for Neuroscience conference. Neural Mechanisms of Afterimages (Sharif Kronemer, Burak Akin, Catherine Walsh, Plyfaa Suwanamalik-Murphy, Cassie Levesque, Javier Gonzalez-Castillo, Daniel Handwerker) This project aims to understand the neural mechanisms of afterimages – the illusory perception of vision following exposure to light. Using high-field fMRI, the goal of this study is to compare whole brain and layer-dependent cortical activity associated with afterimages and images. Our results are summarized by four main findings: (1) afterimages engage widespread cortical and subcortical activity, including across the visual sensory network; (2) the magnitude of the afterimage fMRI signal is weaker in visual sensory regions relative to images; (3) the magnitude of the afterimage fMRI signal is greater in salience network regions relative to images; and (4) afterimages involve greater activity in deeper regions of V1 cortical layers relative to images. This research has been previously presented at the annual meetings for the Association for the Scientific Study of Consciousness (ASSC), Organization for Human Brain Mapping (OHBM), and Society for Neuroscience (SfN). The manuscript highlighting the results from this study is under review at a major journal and we anticipate will be published within the next few months. Visual Perception and Neural Processing of Patients with Cortical Blindness (Sharif Kronemer, Burak Akin, Catherine Walsh, Plyfaa Suwanamalik-Murphy, Cassie Levesque, Javier Gonzalez-Castillo, Daniel Handwerker) This project aims to study residual visual perception and neural processing in patients with cortical blindness – vision loss due to damage in the primary visual pathway. In a recent review article (Kronemer et al., Nature Reviews Neuroscience, 2025), we highlight how physiological measures such as pupil size and skin conductance can infer perception. Therefore, we aimed to investigate how eye measures could detect perception and neural processing in cerebral blindness. We tested people with cerebral blindness to complete a visual perception task with concurrent pupillometry and eye tracking. Images were shown in the intact and blind visual fields. We found that blind field, image-evoked pupil size, blinking, and microsaccades responses were often but not always present for patients with residual blind field conscious awareness. Furthermore, using MEG recordings, we show how these eye metrics are suggestive of blind field visual neural processing. Together, these results support the application of eye metrics and possibly other physiological measures to disclose conscious awareness and neural processing in cerebral blindness. This research has been previously presented at the annual meetings for ASSC and SfN. A manuscript detailing these results is under review at a major journal and we anticipate will be published within the next month.